Each day I attempt to read scientific articles for at least 15 minutes in fields other than behavior analysis. I picked this up following a conversation with a dear friend, Mark Malady. I had just moved to Orlando, FL to begin the master’s program in Applied Behavior Analysis at Florida Institute of Technology and found myself overwhelmed with the amount of knowledge my peers and mentors had obtained about behavior. I immediately began consuming as much behavioral literature as I could. Mark pointed out that, although there was a vast sea of resources to enjoy in behavior analysis, there were also dozens of other relevant fields that would be beneficial for me to at least be somewhat aware of (I now realize that a behavior analyst would benefit from being fluent in these fields).
About six months ago, an article popped up on my news feed titled “Scientists Translate Dolphin Whistles,” in which Janet Fang of I Fucking Love Science (IFLS) reported the findings of researchers in biocommunication and related fields. The researchers created a translator called Cetacean Hearing and Telemetry (CHAT) which utilizes pattern-discovery algorithms “designed to analyse dolphin whistles and extract meaningful features that a person might miss or not think to look for” (Hodson, 2014). Hodson states that the researchers:
detected a whistle for sargassum, or seaweed, which she [the primary researcher] and her team had invented to use when playing with the dolphin pod. They hoped the dolphins would adopt the whistles, which are easy to distinguish from their own natural whistles – and they were not disappointed. When the computer picked up the sargassum whistle, Herzing heard her own recorded voice saying the word into her ear.
The purpose of this blog is to break down the implications of this claim and focus on the benefits and drawbacks of such a technology.
H. Pronko’s From AI to Zeitgeist: A Philosophical Guide for the Skeptical Psychologist
In From AI to Zeitgeist (1988), Nicholas Henry Pronko offers a preliminary examination of the consequences arising from different assumptions in the pursuit of understanding behavior. His book is comprised of a series of topics in alphabetical order with the goal of outlining a different view of the problems, questions, and quandaries bearing on the philosophy of science for psychology (Book FAQ). I’ll attempt to use the ideas discussed in AI to Zeitgeist as I understood the book in an effort to offer a skeptical account of CHAT as well as push the boundaries of my analytical writing skills.
The first idea that comes to mind after reading such an article is, “What cool technology!” The technological advancements of science have clearly benefited our species at an astonishing rate. The thought that we are approaching the technology necessary to speak with another species appears to be very exciting. However, in an effort to provide a skeptical account of such research I would like to pose a question: Is this investigation worthwhile?
Arguments supporting the worthiness of such an investigation may include that communication with other species can lead to a better understanding of their behavior, minds, attitudes, feelings, etc. Although this may seem like a convincing argument, there is little agreement amongst the arts and sciences regarding what constitutes (interspecies) communication. Does interspecies communication involve full comprehension of language? Or is the identification of words via an algorithm adequate? A bigger question might be: are we studying the “language” of dolphins or are we studying how well we impose the English language onto the sounds or “language” of dolphins? Or perhaps even: Are these researchers imposing their own interpretive proclivity for pattern recognition onto the sounds?
Hodson (2014) states that “the team hopes to start trying to figure out what the dolphins’ natural communication means…” The key word here is “means.” Without an explicit definition of “meaning,” it becomes unclear when the researchers have identified a potential answer to their research question. At best it is unclear whether or not this research is answering any questions. In my humble opinion, if scientific evidence cannot convincingly demonstrate reliable measurement and interpretation, this research constitutes nothing more than wasted funding dollars that could have been allocated to research with much more potential for solving large-scale issues (e.g., see the PROSOCIAL movement, or Food Dudes).
My working definition of “meaning” is “the variables responsible for its emission.” This definition enables the identification of behavior-environment relations influencing the emission of a response. As such, any “words” that CHAT identifies through the pattern discovery algorithm would require that the researchers also record the stimulus changes that occur moment-by-moment preceding and following the whistles, not just the kinds of responses emitted by the dolphin. It’s unclear whether CHAT is capable of recording stimulus changes, but it seems unlikely as the relevance of antecedent and consequent events appear to be unconsidered as a matter of the researchers’ philosophic position. Until scientists from varying fields can agree upon terms such as “meaning,” scientific progress (in this case potential behavioral explanations that are objective, parsimonious, and conceptually systematic) are very improbable at best.
An extremely important role of science is replication – it is how we do away with the bunk and woo produced day to day (eventually…). Janet Fang of IFLS points out that “For now at least, the sargassum whistle was just one instance that hasn’t been repeated.” Sidman (1960) elegantly described the importance of direct and systematic replication. With direct replication the same experimenter completes (as close to) the same experimental conditions again with new subjects or repeated observations of the same subjects (Sidman, 1960). Systematic replication is arranging conditions similar to the first experiment, but performing a “new” experiment to obtain additional related data. Sidman states the difference in scientific value with the following statement:
Where direct replication helps to establish generality of a phenomenon among the members of a species, systematic replication can accomplish this and, at the same time, extend its generality over a wide range of different situations.
–Sidman, 1960, p. 111
Simply put, the lack of either systematic or direct replication of “Sargassum” results in the violation of one of the principles of science: replication. Until the findings are replicated with the same researchers (and then in entirely separate research labs – extremely important as well!), the “conclusions” of this research line should be regarded with cautious skepticism.
Big Data & Pattern Discovery Algorithms
Although the invention and implementation of computers has provided the opportunity to create complex representations of phenomenon such as the brain, instead I believe the most significant impact delivered by computers is the ability to perform extremely rapid computations. What used to take a team of researchers days to calculate can now be accomplished in a millisecond. Recent advancements in the area of Big Data and pattern discovery algorithms are particularly interesting to the science of behavior.
The importance placed on single subject research is clearly an important distinguishing characteristic of the behavior analytic tradition. However, as the progression of technology continues to grow exponentially, it may be time to venture outside our comfort zone of N = 1 and venture into the world of N = All. This is what is meant by “big data”: N = All.
Algorithms are sets of rules that a computer follows with the intention of providing an automated problem-solving solution. These algorithms can identify patterns through the use of specific rules. Blending algorithms with the concept of big data may be the method needed to achieve utility in analyzing dolphin sounds. This is because one could potentially collect all of the sounds in a continuous data flow followed by identification of particular patterns. The tricky part is that (a) there may not be a pattern that an algorithm can identify given the fact that the algorithm must be created based on the information that we presently have on hand, and (b) there may not really be a pattern to identify in the first place. We could simply be spending money on searching for an orderly relation that really is not even there!
All of this ties back to my question of whether or not this research is in the best interest of contemporary society. I think one valuable area to study if one chooses to even attempt to answer this question lies in books such as Pronko’s From AI to Zeitgeist (1988). An examination of the consequences arising from different assumptions is much like cleaning your glasses – the world looks a little different…
What are your thoughts? Should this research continue to be funded? Let us know in the comment section below or feel free to email me personally at firstname.lastname@example.org if you would like to continue the discussion.
Sidman, M. (1960). Tactics of scientific research: Evaluating experimental data in psychology. New York, NY: Basic Books. (Reprinted by Authors Cooperative, Boston, MA, 1988)
Ryan O’Donnell is a graduate of Florida Institute of Technology’s applied behavior analysis master’s program. Prior to attending Florida Tech, he received his bachelor’s degree in psychology from the University of Nevada, Reno. His major interests are precision teaching, philosophical positions of the science of behavior, dissemination of behavior analysis, successful applications of technology to increase the efficiency of behavior analysts, and large-scale practical applications of behavioral technology. He works as a Board-Certified Behavior Analyst (BCBA) at High Sierra Industries where he oversees the implementation of behavioral assessment and treatment to adults with a variety of disabilities in a day program.